David, I think I had come across with your library because the link was marked as already visited by my browser :)
Even though I found your code very attractive, what I want to do is to interpolate arbritarily-scattered data, not only points located in a rectangular grid. I managed to perform that task by using the idea behind the shape functions in the finite element method, which gives pretty good results. As a matter of fact, today I took a look at a recommendation of one of my colleagues, that is to arrange the arbitrarily-scattered data into k- dimensional trees to efficiently perform the nearest-neighbor interpolation. I think tha an n-nearest-neighbor or a range-based nearest-neighbor search may also be used to compute some inverse-square weighted sum of individual points or to use some other scheme that provides continuous outputs. I think, at least taking into account only my personal experience with the subject, that multidimensional interpolation has too many caveats and that there may be other aspects that may have more importance to be discussed and/or implemented in GSL. However, it may be handy to have a look at this issue, or even at k-dimensional trees as these artifacts may have other applications as well. -- jeremy On Sunday 05 May 2013 01:42:55 David Zaslavsky wrote: > I have a library that performs 2D interpolation on Github: > https://github.com/diazona/interp2d > It's not ready to be fully released, primarily because there is little > to no documentation or test cases, but I have used the bilinear and > bicubic interpolation routines in a real-world application or two and > they seem to give accurate results. So if you need 2D interpolation, you > could clone the repository and take your chances with it. > > :) David
